An electrically controlled high-speed all-fiber switch is investigated. It is based on a monolithic Mach-Zehnder interferometer using a Gemini fiber. The fiber is provided with internal electrodes for active control of the phase using high-voltage electrical pulses. The demonstrated switching speed is 20 ns. The monolithic design guarantees that the off- and on-states are attained simultaneously for a broad range of wavelengths (50 nm). The interferometer can be switched-off using a second electrode, providing a 15 ns long optical pulse.
©2012 Optical Society of America
Fiber switches are widely used in photonics, where the applications transcend optical communications and include reading different channels in fiber sensor systems, gating fiber lasers or adjusting the delay in switchable recirculation fiber loops. Various types of technologies are employed for switching. All-optical switching can be made very fast, exploiting the weak optical Kerr effects in fibers with intense femtosecond laser [1,2] or fiber resonators that enhance the nonlinear effect [3,4]. If electrical control is desired, low-cost millisecond response-time devices based on fiber pigtailed micro-electro-mechanical-system (MEMS) switches rely on realigning micrometer-size mirrors . MEMS devices have spectrally flat response and small footprint, but have limitations in power-handling capability and speed. Thermooptic planar waveguide circuits can be employed for multiple multiple cross-linked switches with millisecond response time . Electro-optic fiber-coupled lithium niobate Mach-Zehnder interferometers (MZI) provide for picosecond response times , but are costly and suffer from strong polarization dependence. Switching times in the intermediate range (sub-microsecond) is provided for by acousto-optical modulators . Together with the previous technologies, however, pigtailing these devices is not problem-free, introducing loss and limiting the damage threshold. Although successful high-speed all-fiber modulation withstanding the high intracavity power of fiber lasers has been reported recently [9–13], these are often one output-port devices. All-fiber 2x2 Mach-Zehnder modulators have also been reported [14,15], but broadband operation requires that the difference ΔL between the meter-long optical path lengths of the two interferometer arms is very small. The periodicity of a MZI is given by Δλ = λ2/(nΔL), where λ is the optical wavelength, and n is the refractive index. In order to obtain a period larger than 100 nm, ΔL needs to be less than 15 µm, which can be challenging in meter-long devices.
Useful all-fiber interferometers are not easy to implement. As the refractive index depends on both the temperature and the mechanical stresses in the fiber, it is difficult to maintain a stable operation of an interferometer with long independent arms. Recently, a monolithic MZI was described with Gemini fiber which shows enhanced stability and broadband operation . There, passive add-drop multiplexing was demonstrated. In the present paper, an electrically driven, all-fiber 2x2 Mach-Zehnder switch is reported. Because of its monolithic fiber design, it combines useful characteristics of stability, high-power handling capability, nanosecond switching and wide bandwidth in the optical domain.
2. Device fabrication
The monolithic fiber interferometer is based on a combination of the Gemini fiber and an electrically controlled 8-hole Gemini fiber (G2H8), as seen in Fig. 1(a) and (b) . To simplify input coupling, the refractive index profile and dimensions of the these fibers are made to match those of two adjoining standard telecommunication fibers (STF, Corning SMF28), i.e. 0.12 in numerical aperture, 8 µm core diameter and 2x125 µm major axis as seen in Fig. 1(c). Thus, the core separation is 125 µm, which enables splicing between any combination of two adjacent STF, the Gemini fiber and the G2H8. In practice, the active diameter control of the drawing tower was switched off when the Gemini fiber was fabricated due to the difficulty in measuring the size of the fiber in side-view, as the figure-of-eight geometry rotates during drawing. This caused the dimensions of the G2H8 to vary between 125 and 140 µm over the length of the fiber. In order to achieve better mode-matching and lower splice losses, the G2H8 fiber was linearly tapered down to match the 125 µm of the Gemini fiber, a procedure that induced some excess loss.
The schematic layout of the MZI, seen in Fig. 2 , shows that the component consists of two sections of Gemini fiber in which 3-dB couplers are fabricated and a section of G2H8 fiber in between, providing the phase control. Since the fabrication is not automated, all components are made separately and spliced together using a Vytran GPX-3400 processing unit, which enables accurate rotational and transversal alignment along with high-precision tapering. Finally, two adjacent STF are spliced to the input ports of the interferometer in a single operation and the same is done for the output ports. The splicing is carried out with a 1-mm wide graphite filament using several (up to four) 2 s long pulses of 52 W to prevent excessive melting of the fiber tip, which would cause reshaping of the figure-of-eight structure, making simultaneous dual-core alignment difficult while still ensuring a sufficiently strong splice.
The 3 dB couplers are made from the Gemini fiber using the fused biconic taper technique in the GPX-3400 following the procedure described in detail in . The optical cross-coupling is monitored in real-time using a laser source injected into one arm and measuring the optical power at both output arms, allowing the taper process to be halted once the coupling ratio reaches 50%. When the coupler is completed, it is extracted from the machine and glued onto a glass substrate using UV-curing glue. The excess loss of the two couplers is <0.5 dB each. The coupler fabrication process used here is not optimized to make it wavelength insensitive, and this results in a wavelength dependence of ~0.5 dB over the C-band (1530-1565 nm).
The phase modulator in the interferometer is made from a piece of the G2H8 fiber provided with metal electrodes. The holes of the G2H8 fiber are 25-µm in diameter and are located 16 µm edge-to-edge from the core, i.e., sufficiently far to avoid significant metal-induced loss. The holes are filled with Bi-Sn metal and contacted according to the procedure described in . The electrodes are 7-cm long (50 Ω) and the active fiber piece is covered with heat-conductive paste, in order to increase dissipation and allow for higher repetition rates. Although the devices here are made sufficiently long to allow for several trials-and-errors with tapering and splicing parameters leading to an interferometer length of ~4 m, ultimately the interferometer could be constructed to a total length of less than 25 cm, incorporating two couplers and the phase controller.
Two different components are studied, the first one where only a single electrode is contacted (electrode 1) exhibiting a series resistance of 48 Ω, and the second one where two different electrodes are connected, one in each fiber branch (electrodes 1 and 8), with resistances 51 Ω and 54 Ω, respectively. The optical insertion loss of the first components is approximately 8 dB for both outputs. For the second component the insertion loss is approximately 10.5 and 8 dB for the s13 and s14 output respectively. This high excess loss is the combination of four imperfect splices, two tapers, two non-optimized couplers and the metal electrodes.
3. Device operation
The transmission of the two MZIs is measured using an ANDO AQ4321A tunable light source (TLS) and an ANDO AQ6317B optical spectrum analyzer (OSA). The transmission spectrum for the dual-electrode component is shown in Fig. 3 .
The modulation depth is approximately 13 dB, similar to that of the first interferometer which only had a single electrode contacted, and limited by the accuracy of the couplers. The slope in Fig. 3 suggests that the interference period exceeds 200 nm, which means that the optical path length difference is well below 7 µm. The high-frequency oscillations in Fig. 3 are most likely caused by multimode interference between the core and cladding modes stemming from the mode-field mismatch in the splices between the tapered G2H8 and the fiber couplers.
The components were driven by a CMOS circuit, which produces 46 ns pulses of 1.5 kV amplitude, as shown in Fig. 4 (black trace). For the second device, the electrical pulse is divided in a tee-connector to produce two replicas of 1.12 kV amplitude. One of these is delayed by 20 ns using a 4 m-long cable (red and green traces in Fig. 4). A small amplitude reduction is caused by the impedance mismatch, since 45% of the electrical energy exits each one of the two 50 Ω output cables while ~10% is reflected back to the source.
When the electrical current is run through the electrode, resistive heating causes thermal expansion of the metal and induces a pressure increase in the fiber. Due to the photo-elastic and thermo-optic effects, the refractive index changes and the phase of the light in that interferometer arm is altered. If the electrode is heated rapidly, the thermal expansion induces a pressure wave which propagates transversely through the fiber at the speed of sound, which in silica is ~5.96 µm/ns.
The deposited heat eventually dissipates away from the electrode, heats the entire metal-filled fiber by a few degrees (≤10 K), and increases the refractive index through the thermo-optic effect. The thermal effect is much slower than the acoustic effect and only reaches the core within microseconds, compared to the nanoseconds required for the pressure wave . This is illustrated and discussed below, in connection to simulations and Figs. 5 and 6 .
If an identical pulse is applied to the second electrode after a time delay comparable to the time-scale of the acoustic effects, the optical phase-change induced by the first pulse is cancelled by the effects of the second pulse. Thus, the phase shift is turned-off after the chosen delay, with a nanosecond switch-off time. The thermal effects, which are present on a microsecond scale, are also cancelled. Since the same amount of heat is deposited in both branches of the fiber at approximately the same time (on a microsecond time-scale), no residual difference in the refractive index of the two arms is expected.
Simulations of the temperature changes and the variations of the refractive indices for both polarizations are presented in Fig. 5, where COMSOL Multiphysics is used for the calculations. The simulations are performed assuming the application of two 1.12 kV, 46 ns pulses with 20 ns delay to electrodes 1 and 8. The fiber is taken to be encased in a 3 mm diameter cylinder of silicone rubber to mimic the heat conductive paste which covers the component. The critical parameters assumed for the three different materials [silica glass, Bi-Sn, silicone rubber] used in the simulation are : density ρ = [2200, 8560, 1280] kg m−3, thermal expansion coefficient α = [5.5 × 10−7, 15.35 × 10−6, 8.1 × 10−6] K−1, Young’s modulus E = [71.7 × 109, 42 × 109, 4.19 × 106] Pa, Poisson’s ratio ν = [0.17, 0.4, 0.038], thermal conductivity k = [1.38, 19, 0.26] W m−1K−1 and heat capacity Cp = [736, 167, 1803] J kg−1K−1. Additionally, for the silica glass, the refractive indices for the core and the cladding are assumed to be nclad = 1.444 and ncore = 1.4493. The change in refractive indices is calculated usingFig. 1(b).
Since the electrodes that are activated both lie in the x-direction from the cores, the induced stress is anisotropic, which results in birefringence. Measurements on single metal-filled 2-hole fibers with similar hole diameters and separation from the core  have shown that the amplitude of the index change nx induced for light polarized along the direction of the activated electrode is ~3 times larger than the amplitude of the index change ny induced for the orthogonal polarization, but with opposite sign (i.e., nx decreases while ny increases). It was also shown in  that when the mechanical stress wanes and the thermal effect takes over, both nx and ny increase to almost the same positive value.
The simulation results are depicted in Fig. 5 for both the temperature (top row) and the change in refractive index for both polarizations (middle and bottom rows). The color indicates the sign and amplitude of the index change across the Gemini fiber, and it is seen that the core index initially decreases for x-polarized light and increases for y-polarized light. When heat dominates (e.g., 100 µs), the refractive index increases for both polarizations. It is clear that the temperature increase does not reach the cores until ~16 µs.
However, the difference between the x- and y- polarization is still present at 16 µs, as seen as the different shapes of the regions bordering the heated electrodes. Both color scales have been truncated to better observe the effects at the cores, thus the refractive index and temperature changes close to the activated electrodes are beyond the scales shown (limited to < 35°C and < | ± 30 × 10−6|, respectively). In fact, the maximum temperature in the simulation is 51.5 °C inside the electrodes and the maximum and minimum changes of refractive indices in the vicinity of the electrodes are 250 × 10−6 and −83 × 10−6, respectively. It can be noted that while the mechanical stress is present, birefringence is apparent in each branch. However, there is only a difference between cores before the second pulse has been deployed. After that, the value for either x- or y- polarization is the same in both cores, and the relative phase-shift is switched-off.
By taking the spatial average of the refractive index changes for both x- and y- polarization within the cores, the optical phase change for each polarization in either core is calculated fromFig. 6 along with the phase difference between the cores, on both a short- (Fig. 6(a)) and a long time-scale (Fig. 6(b)).
The simulation shows that within the duration of the on-pulse, the x-polarization switches −0.75 π rad while the y-polarization switches + 0.29 π rad. However, because of the off-pulse, the output only reaches −0.33 π rad for the x-polarization and + 0.13 π rad for the y-polarization. After ~16 µs, as the mechanical effect subsides, heat reaches the core and eventually shifts the phase of the x-polarized light to + 1.18 π rad and the y-polarized light to + 1.33 π rad as the temperature in the cores reaches its maximum value.
The transmitted power of the interferometer depends sinusoidally on the phase difference between the two cores and an estimate of how the optical power varies during the heating of the fiber is obtained by taking the cosine of the simulated phase change. The simulated signals for both x- and y- polarizations when only electrode 1 is activated are shown in Fig. 7 along with measurement of both polarizations. The plot is normalized to the initial values of intensity. Qualitatively, the agreement between experiments and simulations is good. It can be noted that the x-polarization switches more than the y-polarization, as predicted by the simulations, and even a change in slope at ~16 µs is apparent for the x-polarization in the simulations and experiments. The y-polarization light switches beyond the π-phase shift since both the mechanical and thermal effects change the refractive index in the same direction. This is observed as a local minimum at ~90 µs. It should be remarked that the simulations are run without any parameter adjustment. Given the material properties and geometry of the device, the calculations of the thermo-optic and mechanical stress result in curves that reproduce well the experimental results (c.f., Fig. 7), with possible exception of the rate of heat transfer. The difference in the time constants of the thermal effect is most likely caused by an over-simplified thermal model (e.g., perfect contact is assumed between the metal electrode and the glass wall of the fiber hole). A small deviation can also be attributed to the experimental alignment of the polarization, responsible for the optical power for the y-polarization exceeding unity after ~50 µs.
4. Measurements and results
The Gemini fiber switch is characterized with linearly polarized light from the TLS set to various wavelengths within and beyond the C-band. The transmitted signal is measured with an InGaAs detector and recorded by a Tektronix TDS3052 oscilloscope on a fast- (Fig. 8(a) ) and on a slow time-scales (Fig. 8(b)). The polarization is aligned along the x-direction to maximize the amplitude of the high-speed switching.
Switching on a nanosecond time-scale is illustrated for the first interferometer when a single 1.5 kV electrical pulse is applied. As seen in Fig. 8(a), the refractive index changes in the core due to the rapid expansion of the electrode and a phase-shift of 0.7 π rad is achieved. It is determined that full switching would require 1.8 kV amplitude, not available here. The risetime of the optical transition is 20 ns, comparable to the electrical pulse duration. A shorter risetime than the electrical pulse duration is expected from the quadratic dependence of the energy with the voltage and the sinusoidal switching efficiency with applied energy. The optical signal in Fig. 8(a) reaches a relatively constant level after switching. However, the small fluctuation between 100 and 200 ns can become larger if the transition time is resonant with a natural acoustic mode. This effect can be damped with accurate packaging . As reproduced in the simulations, when the acoustic pulse dies out, the signal begins to return to the original state. However, after ~16 µs, the thermal diffusion of the deposited heat in the electrode reaches the core, increasing the refractive index. After ~200 µs, the thermal effect has shifted the phase to a maximum. As the heat dissipates further, the signal eventually returns to its original state with a time constant of ~4.3 ms.
The interference pattern for the second component is also approximately flat over the C-band, which means that both devices can be switched from the off-state to the on-state with the settings unchanged for wavelengths between 1520 and 1570 nm.
By applying a 20-ns delayed replica of the electrical pulse to the other branch of the component with two contacted electrodes, the optical signal can be turned off completely as seen in Fig. 9(a) . The thermal effects are also strongly suppressed, as shown on the longer time scale in Fig. 9(b).
It can be noted that the optical pulse is ~15 ns FWHM and that the signal returns to approximately the original value after the pulse, with some small oscillations which subside after 300 ns. When comparing the thermal effects between single pulses and dual pulses, it is clear that the second pulse counteracts the thermal effects of the first pulse. In the present configuration the maximum optical pulse length that can be switched is a few microseconds. In this case, two different pulses would be deployed separately with the desired delay. With an even longer time interval, the decay of the stress and heat reaching the cores returns the device to the off-state before the arrival of the second electrical pulse, leading to imperfect cancellation.
The maximum repetition rate is limited by the heat dissipation which takes place with a ~4.3 ms time constant. Even though the second pulse cancels the interferometric effects of the first pulse, the heat is still present in the fiber and needs to be removed in order to avoid accumulating heat and melting the electrodes. Given the melting point (137 °C), heat capacity (167 J kg−1 K−1) for Bi-Sn and the geometry of the electrodes, the maximum energy that can be deposited by the electrical pulses without damaging the component is ~6.5 mJ. Since a single pulse only delivers ~2 mJ, the temperature increase here is only ~30 °C. If the device is operated at an elevated temperature but still low enough so that the applied energy does not damage the electrode, e.g. around 100 °C, the heat dissipation rate is increased compared to operating at room temperature. Thus, the repetition rate could be significantly increased . However, the accumulated heat should not melt the electrode, and for higher operating rates, other metals such as pure Bismuth (melting temperature 270 °C) may be advantageous.
A high-speed all-fiber switch is demonstrated using Gemini fiber. In this proof-of-concept, the electrically controlled monolithic fiber interferometer provides nanosecond switching for the entire C-band for both the cross and bar state. The all-fiber design has the inherent capability of handling high optical power. When applying a high-voltage nanosecond pulse, the interferometer switches with the similar risetime as the electrical pulse through acoustic effects, while thermal effects come into play after a few microseconds. By applying a second, delayed pulse, the switch is turned-off and nearly perfect compensation of the slow thermal effects is observed. In the demonstration here a 15 ns FWHM optical pulse is generated. The physics of switching is studied involving thermo-optical and acousto-optical effects, and simulations reproduce well the experiments without variable adjustment, suggesting that the main mechanisms are identified. Polarization effects and birefringence are also investigated experimentally and numerically. The device suffers from significant insertion loss, due mainly to the difficulty in splicing because of the varying fiber dimensions. The switching amplitude was limited to ~70% by the electronics and by the maximum voltage that could be delivered to the fiber. The repetition rate of the switch is limited by the cooling of the electrode which is governed by heat dissipation.
We gratefully thank M. Eriksson and H. Eriksson-Quist at Acreo Fiberlab for manufacturing the Gemini and G2H8 fiber. Financial support from the Swedish Research Council (VR) is gratefully acknowledged.
References and links
2. B. K. Nayar, N. Finlayson, N. J. Doran, S. T. Davey, D. L. Williams, and J. W. Arkwright, “All-optical switching in a 200-m twin-core fiber nonlinear Mach-Zehnder interferometer,” Opt. Lett. 16(6), 408–410 (1991). [CrossRef] [PubMed]
4. I. V. Kabakova, B. Corcoran, J. A. Bolger, M. C. de Sterke, and B. J. Eggleton, “All-optical self-switching in optimized phase-shifted fiber Bragg grating,” Opt. Express 17(7), 5083–5088 (2009). [CrossRef] [PubMed]
5. P. De Dobbelaere, K. Falta, S. Gloeckner, and S. Patra, “Digital MEMS for optical switching,” IEEE Commun. Mag. 40(3), 88–95 (2002). [CrossRef]
6. T. Goh, M. Yasu, K. Hattori, A. Himeno, M. Okuno, and Y. Ohmori, “Low loss and high extinction ratio strictly nonblocking 16 x 16 thermooptic matrix switch on 6-in wafer using silica-based planar lightwave circuit technology,” J. Lightwave Technol. 19(3), 371–379 (2001). [CrossRef]
7. G. K. Gopalakrishnan, W. K. Burns, R. W. McElhanon, C. H. Bulmer, and A. S. Greenblatt, “Performance and modeling of broadband LiNbO3 traveling wave optical intensity modulators,” J. Lightwave Technol. 12(10), 1807–1819 (1994). [CrossRef]
9. M. V. Andrés, J. L. Cruz, A. Díez, P. Pérez-Millán, and M. Delgado-Pinar, “Actively Q-switched all-fiber lasers,” Laser Phys. Lett. 5(2), 93–99 (2008). [CrossRef]
10. M. Bello-Jiménez, C. Cuadrado-Laborde, D. Sáez-Rodríguez, A. Diez, J. L. Cruz, and M. V. Andrés, “Actively mode-locked fiber ring laser by intermodal acousto-optic modulation,” Opt. Lett. 35(22), 3781–3783 (2010). [CrossRef] [PubMed]
12. M. Malmström, W. Margulis, O. Tarasenko, V. Pasiskevicius, and F. Laurell, “Soliton generation from an actively mode-locked fiber laser incorporating an electro-optic fiber modulator,” Opt. Express 20(3), 2905–2910 (2012). [CrossRef] [PubMed]
13. J. Li, N. Myrén, W. Margulis, B. Ortega, G. Puerto, D. Pastor, J. Capmany, M. Belmonte, and V. Pruneri, “Systems measurements of 2x2 poled fiber switch,” IEEE Photon. Technol. Lett. 17(12), 2571–2573 (2005). [CrossRef]
14. A. C. Boucouvalas and G. Georgiou, “Fibre-optic interferometric tunable switch using the thermo-optic effect,” Electron. Lett. 21, 512–514 (1985).
15. R. Bahuguna, M. Mina, and R. J. Weber, “Mach-Zehnder interferometric switch utilizing Faraday rotation,” IEEE Trans. Magn. 43(6), 2680–2682 (2007). [CrossRef]
16. P. Rugeland, C. Sterner, and W. Margulis, “Monolithic interferometers using Gemini fiber,” IEEE Photon. Technol. Lett. 23(14), 1001–1003 (2011). [CrossRef]
17. W. Margulis, Z. Yu, M. Malmström, P. Rugeland, H. Knape, and O. Tarasenko, “High-speed electrical switching in optical fibers,” Appl. Opt. 50(25), E65–E75 (2011). [CrossRef]